Although low-loss optical fiber, reliable laser transmitters and low-noise optical receivers have been commercially deployed in fiber optic transmission systems since the early 1980s, it was not until the deployment of reliable commercial optical amplifiers in the early 1990s that the high capacity and low cost-per-bit of wavelength-division multiplexed (WDM) transmission systems could be realized. The introduction of the erbium-doped fiber amplifier (EDFA) revolutionized optical communications by simultaneously amplifying a multiplicity of WDM channels. A single N-channel EDFA replaced N costly regenerators, each composed of an optical-to-electrical converter, a re-timing circuit, a decision circuit, an electronic amplification circuit, and a laser transmitter.
Surprisingly, more than 10 years after its commercial introduction, the EDFA remains the only broadly deployed optical amplifier technology. This, despite the fact that the EDFA has a limited optical bandwidth (under 40 nm) and generally operates only in the C-band (1530-1560-nm wavelength). Although more versatile amplifiers have been demonstrated in the lab, such as semiconductor optical amplifiers (SOA) and Raman amplifiers, those technologies have not gained commercial acceptance due to their inability to compete with the low cost and high performance of EDFAs.
Semiconductor optical amplifiers (SOA) are based on the same mature processing technology as commercial semiconductor lasers. They are far more compact than EDFAs, have the potential for significant cost advantages relative to EDFAs, and have long been considered a natural candidate for use as line amplifiers in WDM transmission systems.
Demonstrations of SOA-based amplified transmission systems predate the invention of the EDFA. SOAs, however, have failed to gain acceptance as a viable WDM amplifier for several reasons, including:
1) Commercial EDFAs can be produced with higher gains and higher output powers than can SOAs.
2) Producing amplifiers with low polarization-dependent gain (PDG) is more challenging for SOAs than for EDFAs.
3) Due to the differences in carrier lifetimes in the two gain media, SOAs are susceptible to inter-channel gain-saturation-induced crosstalk, while EDFAs are largely immune to that adverse effect.
4) SOAs suffer from pronounced gain variations across optical bandwidths typical of WDM systems, while EDFA technology is amenable to the production of multi-stage amplifiers which can incorporate gain-flattening filters and other devices to improve amplifier performance without adversely affecting the amplifier noise figure.
Over the past several years that situation has begun to turn around. Processing improvements have resulted in SOAs with higher gains, higher output powers, and lower PDG. Although modem EDFAs still outperform the latest generation of SOAs, network designers are reconsidering SOAs as the demand for optical networking solutions has increasingly moved from the highly aggregated core network (where performance considerations outweigh cost issues) to the edge of the network (where high performance is often a lower priority than cost and flexibility).
Just as SOA technology was evolving with an eye toward reducing the cost of longer-reach dense wavelength-division multiplexed (DWDM) offerings for this cost-sensitive new market, coarse wavelength division multiplexing (CWDM)—an entirely new un-amplified WDM standard—arose as a low-cost short-reach solution. With optical channel spacings 25-50 times those of standard DWDM, CWDM trades off reach and capacity for cost. CWDM is presently considered an un-amplified technology due to the unavailability of amplifiers that can operate over the wide optical band of a typical system (1470-1610 nm for an 8-channel system). SOAs, which can be engineered to operate at any wavelength within the 1300-1650 nm low-loss region of optical fiber, and typically have a broader band of operation than EDFAs, have been proposed as amplifiers capable of both extending the reach of CWDM systems and reducing the cost of metro DWDM systems (while increasing their flexibility). In the case of CWDM, the SOA gain bandwidth is sufficient to amplify 4 CWDM channels simultaneously. Unfortunately, the best of today's commercially available SOAs still exhibit significant gain variations across that operating band, rendering them unsuitable for most practical CWDM (and DWDM) applications.
The inventors have demonstrated that by combining an SOA with a moderate-gain distributed Raman amplifier, the SOA gain variation can be significantly reduced, the net optical bandwidth can be increased, and the net gain can be increased. Because the Raman gain spectrum in silica fiber has a “ramp” shape, increasing with wavelength until it peaks at roughly 100 nm to the long-wavelength side of the Raman pump wavelength, it is well suited to compensating the monotonic decrease in gain to the long-wavelength side of the SOA gain peak. The SOA provides the bulk of the gain while the Raman gain provides gain tilt compensation. That configuration has an additional benefit in that the information-bearing signals may be positioned to the long-wavelength side of the SOA gain peak (where saturated output powers are highest), resulting in reduced cross-gain modulation in the SOA. The distributed hybrid amplifier 100, shown schematically in FIG. 1, consists of an SOA 140, a Raman pump laser 120, a pump coupler 130, an optical isolator 150, and a long length of single-mode optical fiber 110 that serves as both the transmission fiber and the Raman gain medium.
The amplifier 100 is said to be a “distributed” amplifier because the Raman gain is distributed along the length of the transmission fiber 110. In FIG. 2 is shown a plot 200 of the gain vs. wavelength of the hybrid amplifier and its constituent stages. The triangles 230 represent operation of the SOA alone and show a gain tilt of 6.6 dB over a CWDM 4-channel band 210 between 1510 nm and 1570 nm. The circles 220 represent the Raman gain (increasing by 5.7 dB from 1510 to 1570 nm) achieved by counter-propagating a commercial tunable fiber Raman laser supplying 300 mW of pump power at 1475 nm wavelength through 80 km of reduced water peak standard single mode fiber. That pump wavelength was chosen to produce a Raman gain spectrum which peaks at approximately 1575 nm. The hybrid combination, represented by the squares 250, shows the compensated SOA gain tilt with nearly flat response (0.9 dB variation) over the four CWDM channels with a minimum gain of 16.1 dB. The Raman pump power was intentionally limited to 300 mW to stay within the range of readily available and relatively inexpensive commercial semiconductor Raman pumps.
The data in FIG. 2 represents one instantiation of the above-described distributed amplifier design. With proper choice of SOA and Raman pump wavelength, similar results could have been demonstrated over any band within the wavelength range 1300-1650 nm.
There have been several recent demonstrations of “discrete” optical amplifiers that display wider optical bandwidth, but at a substantial increase in cost. A “discrete” optical amplifier, as opposed to a “distributed” optical amplifier, does not utilize the transmission fiber for amplification. Those broadband amplifier demonstrations have taken advantage of doped fiber amplification (the physical mechanism behind the EDFA) and discrete Raman amplification (using a shorter section of dedicated, highly non-linear fiber (HNLF)).
As mentioned previously, erbium-doped fiber amplifiers (EDFA) are nearly ideal for optical communication applications, due to their maturity, low multi-channel crosstalk, high gain, high output powers, and gain flatness. Unfortunately, commonly available C-band EDFAs cover less than two CWDM channels, and do not have the versatility to operate over any band within the low-loss window of optical fiber.
For example, a new hybrid doped-fiber amplifier is described in T. Sakamoto et al., “Rare-Earth-Doped Fiber Amplifier for Eight-Channel CWDM Transmission Systems,” Optical Fiber Communication Conference and Exhibit 2004, Los Angeles, Calif., paper ThJ5 (March 2004). That amplifier combines three doped fiber amplifier technologies in two branches. One branch consists of an EDFA in series with a thulium-doped fiber amplifier (TDFA). A second, parallel branch consists of an erbium-doped tellurite fiber amplifier (EDTFA). That arrangement promises impressive performance with relatively flat gain of 22.5 dB over the conventional 8-channel CWDM band from 1463 nm to 1617 nm, but is complex, has a high component count, and relies on immature technology, all of which contribute to a high cost for this approach.
Discrete Raman amplification, which can be tailored to a particular wavelength region and gain bandwidth by judicious choice of pump wavelengths and powers, has been demonstrated over the conventional 8 channel CWDM band by the authors of T. Miyamoto et al., “Highly-Nonlinear-Fiber-Based Discrete Raman Amplifier for CWDM Transmission Systems,” Optical Fiber Comm. Conf. and Exhibit 2003, Atlanta, Ga., paper MF19 (March 2003). Their design, using HNLF to increase conversion efficiency, employed six pump lasers with various wavelengths and powers. One, at 1360 nm, was used in both the forward and reverse directions with powers of 211 mW and 614 mW, respectively. Diode laser pumps with output powers above 500 mW are not only more costly than lower power pumps, but are considered Class IV lasers and therefore require more stringent (and costly) laser safety procedures of both vendors and network operators. The remaining pumps had more moderate powers ranging from 236 mW to 7 mW. That design, while simpler than the rare earth doped fiber amplifier mentioned previously, has a lower net gain of approximately 10 dB and is still considered too complex and costly for metro and access applications.
As noted above, CWDM's low cost relative to dense wavelength division multiplexing (DWDM) is attributed to the fact that the CWDM spectrum is orders of magnitude sparser than a typical DWDM spectrum. The ITU standard for CWDM defines a maximum of 18 wavelength channels with a channel-to-channel wavelength separation of 20 nm. That large channel spacing permits approximately a 13-nm channel bandwidth, which in turn makes possible the use of inexpensive CWDM optics and directly modulated, un-cooled semiconductor laser transmitters. As presently implemented, CWDM must be considered an un-amplified technology since the large wavelength spread occupied by all channels in a typical commercial CWDM system (for example, 153 nm for an 8-channel system) cannot be accommodated by readily available low cost optical amplifiers. That constraint limits the reach of most existing CWDM systems to about 80 km.
There therefore remains a need for a low-cost, broadband optical amplifier to overcome those reach limits of commercial CWDM systems. Such an amplifier would also open up new wavelength bands to existing DWDM techniques.